Vulkan and Nvidia

[Fascinate]

How will this be any different than mantle was? Great idea on paper but gets no traction in the market. How many games have the option to run on mantle, like two?

[Alindra]

How will this be any different than mantle was? Great idea on paper but gets no traction in the market. How many games have the option to run on mantle, like two?

Well, the most obvious difference is that Mantle was created by and supported by a single company, AMD. Think of it more as a proof of concept to demonstrate the power and benefits of a low level, low overhead API. It was the driver to get companies to rethink how they did graphics programming. Despite very narrow use, it was quite successful in its goal - it essentially spawned DirectX 12, Metal, and was donated to the Khronos Group as the seed for Vulkan.

Speaking of, instead of being only backed by a single company, it has the backing of the Khronos group (which includes Nvidia, AMD, and Intel). In addition to OpenGL, it will be supported by Linux and Android. It may not be enough to move companies away from DirectX, but it will give it more staying power than Mantle.

Going forward, the performance benefits of this style of API will be too good for companies to pushing graphics fidelity to pass up - be it DirectX 12 or Vulkan. (Granted, I'm biased towards Vulkan because I program in OpenGL and it's not tied to Microsoft's OS.)

[Baelic]

Isn't Vulkan essentially the new OpenGL?

Though this is makes the question that much more difficult. RX480 or GTX1070?

[Zenny]

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

[Bigvizz]

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

In other words AMD is finally meeting expected performance values as the Fury X was always meant to compete with the 980ti and not the 980. So AMD is playing catch up with nvidia, but not really pulling ahead by any significant margin even with DX 12 or Vulcan.

[Vash The Stampede]

How else are they going to make it work when the hardware to do it does not exist on the cards? Magically add the hardware with a driver update?

They'll just download more ram onto the graphics card.



- - - Updated - - -

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

There's only so much that can be done with OpenGL and DX11. AMD cards have a lot of compute power, which is why a lot of bitcoin farmers prefer them. Without anyway to make use of the Async Compute hardware feature, the AMD cards suffered. I believe AMD was assuming that Microsoft would create DX12 a lot sooner, and planned GCN around it. But for whatever reason, Microsoft wasn't working on it until AMD created Mantle. You can't make good use of the hardware unless a game and api is built around it. Yes DX12/Vulkan brings out the full potential of AMD's cards, but not because of shit OpenGL/DX11 drivers.

You can't expect a game to work properly with Async Compute unless it was written specifically for it. This just can't be done in OpenGL/DX11.

In other words AMD is finally meeting expected performance values as the Fury X was always meant to compete with the 980ti and not the 980. So AMD is playing catch up with nvidia, but not really pulling ahead by any significant margin even with DX 12 or Vulcan.

Hence why Mantle was created. AMD dumped Mantle once DX12 and Vulkan was a thing. It wasn't meant to be a serious API on its own, but a way to push for a new low draw call API.

[Deleted]

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

Hmmm, in this bench the fury X completely blows the 980ti out of the water...

85 fps for the 980ti vs 110 for the fury X under Vulkan @ 1440p. Similar difference at 1080p, at 4k the difference is just 5 fps.

[Remilia]

In other words AMD is finally meeting expected performance values as the Fury X was always meant to compete with the 980ti and not the 980. So AMD is playing catch up with nvidia, but not really pulling ahead by any significant margin even with DX 12 or Vulcan.

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

Don't know about you, but even the benchmark here I'm looking at shows it rolling all over the 980Ti and 1070 and going by performance delta, on par with the 1080.
https://www.computerbase.de/2016-07/...ks-amd-nvidia/
That said, I see the "It's cause it's so shit that they see big gains" is well... tiresome. Oh look it sees bigger gains relative to itself therefore it's not an Nvidia fault, except when the card surpasses the competing one does it make it obvious that it just plain out does it better in that API. Even the one you linked shows a 10% higher FPS than the 980Ti. Reality is Nvidia cards gain near nothing on APIs that are meant to give performance boost and better hardware control. It is a problem for Nvidia if they continually show that they can't gain anything from it.
There is something that a lot of people need to look out for in benchmarks is what settings and drivers. 16.7.2 for AMD has better performance on doom Vulkan than the previous. Then if a review isn't using TSSAA or no AA at all, asynchronous compute actually isn't on. They're working on putting that into all AA settings though. And to note AC is only on on AMD cards.

[Zenny]

Don't know about you, but even the benchmark here I'm looking at shows it rolling all over the 980Ti and 1070 and going by performance delta, on par with the 1080.
https://www.computerbase.de/2016-07/...ks-amd-nvidia/
That said, I see the "It's cause it's so shit that they see big gains" is well... tiresome. Oh look it sees bigger gains relative to itself therefore it's not an Nvidia fault, except when the card surpasses the competing one does it make it obvious that it just plain out does it better in that API. Even the one you linked shows a 10% higher FPS than the 980Ti. Reality is Nvidia cards gain near nothing on APIs that are meant to give performance boost and better hardware control. It is a problem for Nvidia if they continually show that they can't gain anything from it.
There is something that a lot of people need to look out for in benchmarks is what settings and drivers. 16.7.2 for AMD has better performance on doom Vulkan than the previous. Then if a review isn't using TSSAA or no AA at all, asynchronous compute actually isn't on. They're working on putting that into all AA settings though. And to note AC is only on on AMD cards.

That benchmark is so full of shit, while the Fury X is faster then the 980ti in Vulkan the performance difference is not nearly as large:

https://www.youtube.com/watch?v=ZQi6wwSOhZU
http://www.eurogamer.net/articles/di...formance-gains

Yes the Fury X is 7% faster then the Geforce 980ti but that is hardly a massive difference, it's not as if the Geforce cards do worse, they actually get a 5%-10% performance increase. There is also the fact that Nvidia is utilizing a older version of Vulkan when compared to AMD. For some reason.

[Evildeffy]

The Vulkan version is irrelevant.

You have the Engine Version -> Engine Update -> Fix/Change version.

There are 0 performance differences between 1.0.8 and 1.0.17 as it would have to be 1.1 to make any possible meaningful difference.
Because only engine updates would facilitate any form of performance changing code.

Simple fact of the matter is thus:
AMD built their cards around low level advantages because they knew devs were screaming for it and they expected it to happen a lot sooner.
For lack of that API to come out AMD developed Mantle as a stepping stone and it did exactly as it was intended to, shake up MS and kick shit into gear.
AMD thus holds the advantage in any API in low level environments as their cards are specifically designed for it.

nVidia designed the hardware of their cards not around low level but DX11 because that was, and for the moment still is, the primary graphics API.
But nVidia grossly miscalculated how fast DX12/Vulkan was desired by devs and the adoption rate is through the roof.

So in simple terms:
It does not matter how much you would like to say "nVidia supports it" in reality they only support it if wanted and often incur a penalty in performance when enabled.
AMD cards in that sense are flat-out better, this is what makes them better at VR as well because it's built on the same low level requirements.

Does that make nVidia cards bad? No it doesn't, the cards themselves are good cards.
The company is another matter as they try to hide their shortcomings by lying, cheating etc.

Does that make AMD cards worse? No it doesn't, the cards themselves are good cards.
The company is another matter as they had inefficient drivers previously but are working to correct that.

And if you really want to be technical regarding AMD's OpenGL performance...
nVidia ran OpenGL 4.5 with the latest version which also included some massive performance boosts over OpenGL 4.3 which DOOM ran on for AMD.
The funny part about this was that in the Multiplayer Beta both ran OpenGL 4.5 and the AMD cards were dominating over nVidia.

GTX 1080 gets released/showcased and AMD all of a sudden "lost" the capability to run OpenGL 4.5 (even though it's fully compliant) and lost performance.

I would rather say that looking at the history of things it would've been rather even in this game but Vulkan simply utilizes more of the hardware for each brand.
The problem here is that nVidia has barely any added horsepower left where AMD cards have untapped resources ready and waiting.

So in short:
When it concerns low level APIs, which you SHOULD be focusing on as it's adoption rate is insane, AMD is without question the current victor.
If nVidia's Volta ever gets released it might have the hardware to challenge AMD in this but right now they simply cannot physically do it.
That said if Volta gets the design and hardware to support it properly they will suffer a performance penalty in comparison to their older cards as hardware design will change a lot of things and you cannot make something like Maxwell 3.0 (Oh sorry I mean Pascal...) and simply drop in some hardware and be done with it.. nope that will make 0 difference, some things will have to be given up to favour the more powerful API so serial processing will take a hit meaning that if you built an nVidia card on the exact same FinFET with the exact same "IPC" etc. just the difference is serial processing or parallelization and try to run DX11 stuff the Pascal card would do it better than the Volta card just as Volta would run DX12/Vulkan developed games better than Pascal.
Hardware designs make a huge difference and if nVidia adapts to the new way of things and don't muck it up you could have a beast.

So either has the chance to shine for it but right now that spotlight is on AMD with low level APIs.

[Rixis]

How else are they going to make it work when the hardware to do it does not exist on the cards? Magically add the hardware with a driver update?

Reminds me of that time 4chan pushed an advert and managed to convince people the iOS update made their iphone waterproof. There were enough complaints of people dropping their phones in the sink to check that you just have to marvel at marketing and what they can sell you.

[Deleted]

Interestingly I've noticed a higher CPU load on VULKAN compared to Open GL

http://imgur.com/a/AdyeM

[Artorius]

It's not that Nvidia is bad at OpenGL it's that AMD is so shit at it, thus seeing massive performance gains from the switch to Vulkan:

http://www.eurogamer.net/articles/di...formance-gains

Look at that, the Fury X was basically on par with the Geforce 980 under OpenGL, under Vulkan it is on par with the 980ti which it should be.

In other words AMD is finally meeting expected performance values as the Fury X was always meant to compete with the 980ti and not the 980. So AMD is playing catch up with nvidia, but not really pulling ahead by any significant margin even with DX 12 or Vulcan.

The misinformation spreading is real...

Please guys, it's pretty clear that you don't know a thing about hardware design so just don't talk about something that you don't know about. It's simple.

It's obvious that the Fury X can't bring all of its potential using DX11, Fiji XT is a GPU from 2015 designed to take advantage of the things that are relevant now and in the future. It isn't designed around an API from 2009.

Nvidia's GPUs focused on power efficiency and simply don't even have the hardware to do it. They decided that they didn't need it for games because they really didn't some years ago and GPU design goes on for years. They probably will have proper support with their next gen or something, but they don't now and their cards don't show any meaningful performance increase under those new APIs (DX12 and Vulkan). This isn't rocket science.
- - - Updated - - -

Interestingly I've noticed a higher CPU load on VULKAN compared to Open GL

http://imgur.com/a/AdyeM

That's intended.

[Barnabas]

Nvidia doesn't need to be the answer to everything company. They have cards that are good at certain things and their tech shines. AMD has things that make their cards shine.

[Bigvizz]

The misinformation spreading is real...

Please guys, it's pretty clear that you don't know a thing about hardware design so just don't talk about something that you don't know about. It's simple.

It's obvious that the Fury X can't bring all of its potential using DX11, Fiji XT is a GPU from 2015 designed to take advantage of the things that are relevant now and in the future. It isn't designed around an API from 2009.

Nvidia's GPUs focused on power efficiency and simply don't even have the hardware to do it. They decided that they didn't need it for games because they really didn't some years ago and GPU design goes on for years. They probably will have proper support with their next gen or something, but they don't now and their cards don't show any meaningful performance increase under those new APIs (DX12 and Vulkan). This isn't rocket science.

I will continue to have my opinion regardless of yours /shrug
Still..... no fucks given Artorius. Have a nice day.

[Kuntantee]

I will continue to have my opinion regardless of yours /shrug
Still..... no fucks given Artorius. Have a nice day.

So you will insist on your ignorance? Spoken like a true interwebz user.

- - - Updated - - -

Nvidia doesn't need to be the answer to everything company. They have cards that are good at certain things and their tech shines. AMD has things that make their cards shine.

If they don't answer to async compute, Nvidia will lose the gaming market. Async compute seems like where the market is heading.

[Bigvizz]

Nvidia doesn't need to be the answer to everything company. They have cards that are good at certain things and their tech shines. AMD has things that make their cards shine.

I agree with you, Nvidia is better at Marketing and Business, Where I believe AMD has an edge on Hardware, but at the end of the day a Company that knows how to sell their product wins in the end.

- - - Updated - - -

So you will insist on your ignorance? Spoken like a true interwebz user.

I'm sure someone cares, but I don't. lol

[Kuntantee]

I'm sure someone cares, but I don't. lol

You care, more than most people here. You are either emotionally attached to Nvidia, what we call a fanboy, or subliminally angry to the situation because you bought an overpriced Nvidia card which may be beaten by cheaper AMD cards in the near future.

[Drunkenvalley]

Openly admitting you are spreading misinformation but not caring is called trolling, yes?

[Vash The Stampede]

If they don't answer to async compute, Nvidia will lose the gaming market. Async compute seems like where the market is heading.

Considering how Pascal was just released, Nvidia may have a year or two of being the underdog. But Nvidia's advantage is marketing, and influence. Watch as some websites continue to ignore Vulkan and DX12 benchmarks. The 1060 is about to be released and we have a plethora of Vulkan/DX12 games to test it against the RX 480.

- - - Updated - - -

I agree with you, Nvidia is better at Marketing and Business, Where I believe AMD has an edge on Hardware, but at the end of the day a Company that knows how to sell their product wins in the end.

That is something we as consumers should fight against. We want the best product for the lowest price, and Nvidia's Founders Edition is going the wrong way. Right way for Nvidia, just wrong for consumers. Going to see a lot of people upset when their 1080 Founder's Edition $699 graphics card out performed by the $399 Fury X.